Artificial intelligence is behind one of the largest mobilizations of capital in recent history. Major technology companies are investing hundreds of billions of dollars in data centers, semiconductors and energy infrastructure. Meanwhile, financial markets are caught between enthusiasm for a revolution that promises to transform the economy and fear of fueling a new bubble.
The comparison with the dot-com bubble at the turn of the century is inevitable. But there is a fundamental difference. This time, we are not talking only about expectations for digital businesses. We are talking about massive investment in physical assets. AI requires buildings, land, electricity, cooling systems and billions in technology equipment. In a sense, the most digitally driven revolution in our history is also fueling an extraordinary cycle of real estate and industrial investment.
This changes the nature of the risk. A software company can scale with relatively little capital. A data center ties up enormous sums before anyone knows whether it will pay off. And unlike most real estate, much of its equipment can become technologically obsolete within a few years.
The numbers show the scale. In early October, Nvidia came close to a market capitalization of $6 trillion. Only days later, the market began to question the valuations of some companies tied to AI infrastructure. On October 14, Nebius fell 15.1%, CoreWeave 9.5% and Oracle 7.2% in a single session. Notably, even as these companies took sharp corrections, the S&P 500 was still up nearly 13% for the year.
Another factor deserves attention. These investments are not driven by today’s demand alone. They are also driven by a race to secure competitive advantage. Microsoft, Google, Amazon and Meta need computing capacity, but they also need to keep their competitors from controlling it. In that environment, investing can become a strategic necessity, even when financial returns remain uncertain.
Then there is a particularly delicate issue: circular investment. Large technology companies fund startups, which then use the money to buy infrastructure and technology from their own investors. Such arrangements can accelerate the sector’s development. They also make it harder to distinguish genuine demand from demand stimulated by the financing itself.
One thing is the profitability of those who build AI infrastructure, and quite another is the value that this technology can generate for the companies that use it.
The question is no longer simply whether there is an AI bubble. There are enough signs of financial exuberance, potential overinvestment and outsized expectations to take that possibility seriously. The real unknowns are where the excess is concentrated, when a correction could occur and who would bear the consequences. Economic history shows that technological revolutions can coexist with major financial excesses. The internet transformed the world, yet many of the companies that led its first expansion disappeared.
This is where an essential distinction comes in. The profitability of the companies building AI infrastructure is one thing. The value the technology can create for the businesses that use it is quite another. For those businesses, the opportunity should not be measured by Nvidia’s share price or the number of models available. It should be measured by their ability to improve productivity, redesign processes, make decisions faster and transform their operating models.
In fact, a financial correction could make certain technologies cheaper and easier to adopt. The winners of this revolution may not be the companies with the highest market valuations today.
That is why, amid the AI-bubble debate, businesses should ask a different question: not how long the financial enthusiasm for AI will last, but how much real value they can capture from it. Because we may be overvaluing some AI companies while, at the same time, deeply undervaluing AI’s capacity to transform our businesses.
The article was first run by Revista Capital.